The fact that you chose the non node model proves you don't understand what I was talking about.
Your assertion that I don't know what I'm talking about is entirely correct. But I'll continue on anyway.;-)
You're right, the Xserve Node is the logical choice for a cluster and we should be referencing it if we care about any sort of serious comparison of an Xserve versus an Xserve's cost in Mac Mini's. You seem to feel that this is a no-contest win for the Xserve and that the idea of using Mac Minis in a cluster is a joke, correct. But rather than dismiss the idea outright, let's actually look at the numbers:
So for processing power, the Mini cluster takes the lead, although certain situations may exist where the G5s could outperform the G4s as the Xserve has an impressive system bus that gives it additional performance gains not reflected in the above numbers. For memory, the Minis' take a clear and commanding lead in quantity, but they lose in memory access speed. Also additionally unlisted in the above chart is the Xserve's superior memory bandwidth. Moving on to storage space, the Minis win a dramatic hands-down victory, but they then lose out badly in the network bandwidth department. But on the whole, the Minis appear to be standing up well in many areas.
But the tides can turn if you want to max out some of the available build-to-order options. For example, this is what you get if you max-out the RAM in an otherwise stock configuration:
With equal amounts of RAM, slightly better access speeds and superior memory bandwidth, the Xserve now clearly out-performs the Mini in this area.
If we decide we value storage space enough to maximize our HDs in otherwise stock configs, we get these numbers:
(Maxed Storage). 1 XSERVE NODE........ 6 MAC MINIS Price:.......... $3,449............... $3,294 (6x $549) Storage:........ 400GB Serial ATA..... 480GB ATA Ultra (6x 80GB)
Now the Xserve is still behind, but not by much. (For this case, "maxed out storage" only refers to the maximum storage I can BTO from Apple's web site. It should be noted that the Xserve can theoretically top out at 1.2TB. It should also be noted that I'm not bothering to compare to the terrifying capacity of the Xserve RAID. I mean geez.)
Of course, these adjustments to the stock config require us to imagine that we value these specific features that favor the Xserve in a comparison. But depending on our task, we may not actually place such a high value on these things and instead lean back towards the stock config. For example, if we say we value raw computational power over these other considerations, the Mini may still stand up well. After all, it's got GHz on its side, while the Xserve has the stronger architecture and system support. I'd personally love to see real-world task performance comparisons.
In conclusion, after this analysis I hope most will agree that while the Xserve remains a sensible choice, the Mac Mini delivers such surprising value for its size and power that the very idea of a Mac Mini cluster need not necessarily be written off as a joke.
(This whole thread reminds me of the age-old question "Which would win in a fight: A Rottweiler, or a Rottweiler's weight in Chihuahuas?" But that's a comparison for another day.)
Um, yes. I did see that the first line of your post referred to a cluster, but the second line of your post then referred to the submitter, who seemed to be (mainly) linking to a colocation site, and so I mentally started assuming you must be using the word cluster loosely, as "a number of similar things collected together or lying contiguous". I stand corrected.
Back on track now, let's look at your numbers. A 2.3GHz dual G5 Xserve for $2,300 you say? May I ask where you're getting that number? Looking at Apple's web site now, I see that same model going for $3,999. And even the low-end 2.0GHz single-processor G5 Xserve is $2,999. Both models support up to 8GB of memory.
I'm not arguing that Mac Minis' would make a better price/performance cluster, I'm just clarifying the numbers involved for anyone who wishes to embark upon such as argument.
If you're comparing cost per unit of storage or bandwidth (and even perhaps processing power), the Xserve is going to win. But for small businesses even a single Xserve may be excessive. If you instead compare total cost out of pocket, a colocated Mac Mini suddenly looks like a superstar. The colocation service linked to is potentially a great way for a small business with limited product demand to cheaply and reliably serve the internet.
If you want to compare againt cheap PCs instead of Xserves, size will likely be your issue. From what I've seen, all PCs priced cheaper than a Mac Mini are physically larger. Any colocation service would likely charge more for the additional space consumed.
Thanks, minimunchkin! I rarely bother to post to slashdot, but when I do I often think to myself "If I entertain only one person with this post, I'll be worth all those years that I otherwise would have wasted learning English as a first language."
And in a pathetic attempt to remain vaguely on-topic, check out this news. Dell's CEO, in an amazing display of delayed postnatal mental retardation, dismisses the iPod as a "fad". You know, like the Walkman was. We're left to assume that if a product only rakes in unbelievable craploads of cash for a mere 10-15 years, at the end of the day we should ask ourselves "was it worth it?" He then goes on to turn a blind eye to the possibilities Apple is making for itself with the Mac Mini. Check it yourself at http://news.com.com/Dells+Rollins+Unfazed+by+iPod% 2C+IBM/2008-1082_3-5540420.html?tag=nefd.ac before madly rushing to sell your stock in Dell.
The point was supposed to be that anyone who has ever enjoyed a tasty meal can hold values beyond nutrition. But if you've only ever had gruel your entire life, you may not understand what you're missing. Now this may or may not apply to you; although you've stated that you didn't buy an iPod, I am not yet psychic enough to accurately intuit your level of experience with said device. But what I can observe is that you only seem to require the most rudimentary functionality from your product. Saying that you only require your MP3 player to play music is like requiring your drinks to be wet or your legs to reach the ground.
The Creative does everything my wife wants it to do... which is basically play music. Nothing more, nothing less. It shows the song info on the screen for each song and the random works fine. What else do you really need from an mp3 player? Sure I guess it might be nice if it could grill me a steak but it was purchased to do only one thing, and to do that one thing well. It does that. What else could/should we be asking of it?
That's like saying "Our food does everything my wife wants it to do... which is basically to provide sustenance. Nothing more, nothing less. It is nourishing and can be penetrated with a fork. What else do you really need from a meal? Sure I guess it might be nice if it could [insert random unrelated task here] but it was purchased to do only one thing, and to do that one thing well. It does that. What else could/should we be asking of it?"
But hey, if it works for you, I won't knock it. Bliss is in the eye of the beholder, or something like that.
Of course they should be randomly audited, but against what? Exit polls and past voting demographics by precinct?
This is why electronic voting is such a terrible idea. There is no good way to audit the votes. If you have a reciept that you drop in a box, why not just use paper. If you have some sort of personal identifier, why not just advertise your vote on some website?
Random audits will work if we adopt a voter-verified paper trail. We don't need to be afraid of electronic voting if we just commit to doing it right. Compare our system to Australia's for a good example. http://www.wired.com/news/ebiz/0,1272,61045,00.htm l
And if we do finally get it right and we establish an electronic voting system that we trust, it will offer a clear advantage over paper ballots. And I'm not just talking about speedy results. For example, some machines allow visually impaired voters to cast their vote without human assistance. Keep in mind that while paper ballots may be great in your county (we use an excellent, clearly formatted optical-scan ballot in mine), many paper ballots suffer from problems such as hanging chads in punch-card systems, or confusing candidate layouts. Remember Florida's butterfly ballot of 2000? http://news.bbc.co.uk/1/hi/in_depth/americas/2000/ us_elections/glossary/a-b/1037172.stm
I don't know about you, but when I voted i signed in and they took my name as having voted. There's your audit trail.
But this audit trail will only tell us the total number of votes, not the number cast for each individual candidate. This is an inadequate audit trail because it will fail to catch cases where one candidate's votes are increased by the same number another candidate's votes are decremented. And even when it does manage to catch a problem where the vote total recorded by the machine does not match the vote total listed on paper, there's no way to determine what the correct totals for each candidate should be.
To have a truly meaningful audit trail we need each electronic vote cast to be mirrored by a voter-verified physical receipt. Simply knowing what the combined vote total should be doesn't tell us which candidate should have won. And if we're serious about practicing democracy, we can't settle for anything less.
I'm confused. What is the advantage of paper? Everyone keeps saying "OH, it's necessary," but why?
Currently, without a paper trail, voters must trust the machine to record their vote correctly. The machine could swear up and down that it recorded your vote for candidate X, but how do you know it didn't (accidentally or maliciously) secretly cast it for candidate Y? The vote is recorded on a memory card that the voter cannot view personally. And even if they could, they probably wouldn't understand which ones and zeros represented their vote or how it was cast. And if the numbers the machine spits out look odd, there's no meaningful way to conduct a recount.
But we can change this by adding a paper trail. Imagine if each voting machine, in addition to recording the vote in memory, also presented a receipt which stated in plain english how the machine just voted. The voter need never even touch this receipt; it could be provided behind glass. After reading the receipt and verifying that it recorded the vote correctly, the voter could then press a button to indicate their approval and the receipt could be automatically dropped into a secure ballot box. This doesn't prevent the machine from recording the vote incorrectly in memory, but it allows for the results to be audited. At the end of the day, if warranted, authorized poll workers would be able to manually count all the receipts and compare them to the numbers the machine recorded, thus allowing for meaningful recounts.
Would you use a bank that didn't send you statements or give you receipts from the ATM? "Trust us," they tell you. "We're handling your money just fine. After all, we're using machines!" Would such a statement inspire confidence in your financial institution? Or would you worry about errors or possibly fraud?
Paper trails don't prevent problems, but they allow for audits so that problems can be caught and corrected.
I suppose then that I would disagree that we need to remove strategic voting. In theory (Duyba doesn't support this theory but Clinton did) it should pull the politician towards the center to appeal to more voters.
Strategic voting doesn't do anything of the sort. I believe you're confusing voting strategy with campaigning strategy, and further assuming that a winning campaign strategy should be for a candidate to appeal to the majority of voters instead of their base. And that's a risky assumption.
I further suspect that you may not really understand the issue at hand because your comment states both a belief that we need not remove the strategic aspect of voting (an opinion that necessarily allows for the possibility of such removal) and a belief that all voting is strategic (a belief that necessarily disallows the possibility of separation). So I will attempt to outline the issue as I understand it so that we may have a common ground for discussion.
When people talk about the need to remove the strategic vote, what they mean is they want to remove or reduce the circumstances under which a voter must choose between voting for the candidate they personally favor (their sincere choice) and a candidate they think might win (their strategic choice). Such situations are common when using First-Past-The-Post (A.K.A. plurality voting) methods where voters casting ballots for their sincere choice often feel that their vote is "being thrown away" if it's going towards a candidate who has no reasonable chance of winning. And so instead they often cast their ballot strategically for a less desirable candidate who is more likely to win. This in turn leads to a less representative government, which is bad, because representative government is the goal of democracy! It's because of this that advocacy of other voting methods that offer no incentive to vote strategically is gradually increasing.
This is a _terrible_ idea. If voters walk out of the voting booth with a receipt that says how they voted...
That is NOT what Voter Verifiable Voting means. VVV simply means that the voter can personally verify that their vote was recorded properly.
Today, with paper ballots, this is as easy as looking at your ballot and making sure you put the mark next to the intended candidate. You don't keep this with you when you leave the polling station; you drop it in the ballot box. Nobody knows how you voted but you.
Today, using electronic voting machines, this does not exist. The machine may tell you that it cast your vote for the candidate you intended, but ultimately the machine records your vote on a memory card that you don't have access to. In other words, the voter has substituted their own verification of the vote for the machine's verification of their vote. And without the voter being able to personally verify that their intended candidate received their vote, in the end, they're forced to trust that the machine has made no error and his recorded the vote honestly. Many voters do not have this trust. And with the myriad accounts of voting machine errors and tampering in circulation today, rightly so.
So when people loudly decry that our voting machines need Voter Verifiable Voting, what they're usually looking for is a paper trail that they can personally see and understand. For example, a slip of paper that says "You voted for John Smith for President". This paper receipt need not be kept by the voter! It could be verified by the voter and then dropped in a secure ballot receipt box. The voter need not even personally touch this receipt! It could be presented behind glass for verification by the voter, to be secured upon the voter's approval. And while the machine would still be used to count and report the votes, the paper receipts would allow for meaningful recounts and would provide a paper trail for audits. These are good, important things.
And while there exists a poorly informed minorty that would like to be able keep this voter verifiable receipt with them when they leave the polls, these people represent a distinct subset of people calling for Voter Verifiable Voting in general. So let's be sure not write off VVV as a whole because there exists a handful of people who want to implement it poorly.
While the Neuros seems to have a lot of cool features, it would appear that Mac users need not apply. Here are the system requirements as taken from their website, for those who care:
OS: Microsoft® Windows 98SE/Me/2000/XP
CPU: Pentium 233MHz or higher
Memory: 64MB minimum
Hard Drive: 160MB
USB Port
Slashdot Mirror
Your assertion that I don't know what I'm talking about is entirely correct. But I'll continue on anyway. ;-)
You're right, the Xserve Node is the logical choice for a cluster and we should be referencing it if we care about any sort of serious comparison of an Xserve versus an Xserve's cost in Mac Mini's. You seem to feel that this is a no-contest win for the Xserve and that the idea of using Mac Minis in a cluster is a joke, correct. But rather than dismiss the idea outright, let's actually look at the numbers:
So for processing power, the Mini cluster takes the lead, although certain situations may exist where the G5s could outperform the G4s as the Xserve has an impressive system bus that gives it additional performance gains not reflected in the above numbers. For memory, the Minis' take a clear and commanding lead in quantity, but they lose in memory access speed. Also additionally unlisted in the above chart is the Xserve's superior memory bandwidth. Moving on to storage space, the Minis win a dramatic hands-down victory, but they then lose out badly in the network bandwidth department. But on the whole, the Minis appear to be standing up well in many areas.
But the tides can turn if you want to max out some of the available build-to-order options. For example, this is what you get if you max-out the RAM in an otherwise stock configuration:
With equal amounts of RAM, slightly better access speeds and superior memory bandwidth, the Xserve now clearly out-performs the Mini in this area.
If we decide we value storage space enough to maximize our HDs in otherwise stock configs, we get these numbers:
Now the Xserve is still behind, but not by much. (For this case, "maxed out storage" only refers to the maximum storage I can BTO from Apple's web site. It should be noted that the Xserve can theoretically top out at 1.2TB. It should also be noted that I'm not bothering to compare to the terrifying capacity of the Xserve RAID. I mean geez.)
Of course, these adjustments to the stock config require us to imagine that we value these specific features that favor the Xserve in a comparison. But depending on our task, we may not actually place such a high value on these things and instead lean back towards the stock config. For example, if we say we value raw computational power over these other considerations, the Mini may still stand up well. After all, it's got GHz on its side, while the Xserve has the stronger architecture and system support. I'd personally love to see real-world task performance comparisons.
In conclusion, after this analysis I hope most will agree that while the Xserve remains a sensible choice, the Mac Mini delivers such surprising value for its size and power that the very idea of a Mac Mini cluster need not necessarily be written off as a joke.
(This whole thread reminds me of the age-old question "Which would win in a fight: A Rottweiler, or a Rottweiler's weight in Chihuahuas?" But that's a comparison for another day.)
Back on track now, let's look at your numbers. A 2.3GHz dual G5 Xserve for $2,300 you say? May I ask where you're getting that number? Looking at Apple's web site now, I see that same model going for $3,999. And even the low-end 2.0GHz single-processor G5 Xserve is $2,999. Both models support up to 8GB of memory.
I'm not arguing that Mac Minis' would make a better price/performance cluster, I'm just clarifying the numbers involved for anyone who wishes to embark upon such as argument.
If you want to compare againt cheap PCs instead of Xserves, size will likely be your issue. From what I've seen, all PCs priced cheaper than a Mac Mini are physically larger. Any colocation service would likely charge more for the additional space consumed.
And in a pathetic attempt to remain vaguely on-topic, check out this news. Dell's CEO, in an amazing display of delayed postnatal mental retardation, dismisses the iPod as a "fad". You know, like the Walkman was. We're left to assume that if a product only rakes in unbelievable craploads of cash for a mere 10-15 years, at the end of the day we should ask ourselves "was it worth it?" He then goes on to turn a blind eye to the possibilities Apple is making for itself with the Mac Mini. Check it yourself at http://news.com.com/Dells+Rollins+Unfazed+by+iPod% 2C+IBM/2008-1082_3-5540420.html?tag=nefd.ac before madly rushing to sell your stock in Dell.
The point was supposed to be that anyone who has ever enjoyed a tasty meal can hold values beyond nutrition. But if you've only ever had gruel your entire life, you may not understand what you're missing. Now this may or may not apply to you; although you've stated that you didn't buy an iPod, I am not yet psychic enough to accurately intuit your level of experience with said device. But what I can observe is that you only seem to require the most rudimentary functionality from your product. Saying that you only require your MP3 player to play music is like requiring your drinks to be wet or your legs to reach the ground.
That's like saying "Our food does everything my wife wants it to do... which is basically to provide sustenance. Nothing more, nothing less. It is nourishing and can be penetrated with a fork. What else do you really need from a meal? Sure I guess it might be nice if it could [insert random unrelated task here] but it was purchased to do only one thing, and to do that one thing well. It does that. What else could/should we be asking of it?"
But hey, if it works for you, I won't knock it. Bliss is in the eye of the beholder, or something like that.
This is why electronic voting is such a terrible idea. There is no good way to audit the votes. If you have a reciept that you drop in a box, why not just use paper. If you have some sort of personal identifier, why not just advertise your vote on some website?
Random audits will work if we adopt a voter-verified paper trail. We don't need to be afraid of electronic voting if we just commit to doing it right. Compare our system to Australia's for a good example. http://www.wired.com/news/ebiz/0,1272,61045,00.htm l
And if we do finally get it right and we establish an electronic voting system that we trust, it will offer a clear advantage over paper ballots. And I'm not just talking about speedy results. For example, some machines allow visually impaired voters to cast their vote without human assistance. Keep in mind that while paper ballots may be great in your county (we use an excellent, clearly formatted optical-scan ballot in mine), many paper ballots suffer from problems such as hanging chads in punch-card systems, or confusing candidate layouts. Remember Florida's butterfly ballot of 2000? http://news.bbc.co.uk/1/hi/in_depth/americas/2000/ us_elections/glossary/a-b/1037172.stm
But this audit trail will only tell us the total number of votes, not the number cast for each individual candidate. This is an inadequate audit trail because it will fail to catch cases where one candidate's votes are increased by the same number another candidate's votes are decremented. And even when it does manage to catch a problem where the vote total recorded by the machine does not match the vote total listed on paper, there's no way to determine what the correct totals for each candidate should be.
To have a truly meaningful audit trail we need each electronic vote cast to be mirrored by a voter-verified physical receipt. Simply knowing what the combined vote total should be doesn't tell us which candidate should have won. And if we're serious about practicing democracy, we can't settle for anything less.
Currently, without a paper trail, voters must trust the machine to record their vote correctly. The machine could swear up and down that it recorded your vote for candidate X, but how do you know it didn't (accidentally or maliciously) secretly cast it for candidate Y? The vote is recorded on a memory card that the voter cannot view personally. And even if they could, they probably wouldn't understand which ones and zeros represented their vote or how it was cast. And if the numbers the machine spits out look odd, there's no meaningful way to conduct a recount.
But we can change this by adding a paper trail. Imagine if each voting machine, in addition to recording the vote in memory, also presented a receipt which stated in plain english how the machine just voted. The voter need never even touch this receipt; it could be provided behind glass. After reading the receipt and verifying that it recorded the vote correctly, the voter could then press a button to indicate their approval and the receipt could be automatically dropped into a secure ballot box. This doesn't prevent the machine from recording the vote incorrectly in memory, but it allows for the results to be audited. At the end of the day, if warranted, authorized poll workers would be able to manually count all the receipts and compare them to the numbers the machine recorded, thus allowing for meaningful recounts.
Would you use a bank that didn't send you statements or give you receipts from the ATM? "Trust us," they tell you. "We're handling your money just fine. After all, we're using machines!" Would such a statement inspire confidence in your financial institution? Or would you worry about errors or possibly fraud?
Paper trails don't prevent problems, but they allow for audits so that problems can be caught and corrected.
Our democracy is important. Let's act like it.
Strategic voting doesn't do anything of the sort. I believe you're confusing voting strategy with campaigning strategy, and further assuming that a winning campaign strategy should be for a candidate to appeal to the majority of voters instead of their base. And that's a risky assumption.
I further suspect that you may not really understand the issue at hand because your comment states both a belief that we need not remove the strategic aspect of voting (an opinion that necessarily allows for the possibility of such removal) and a belief that all voting is strategic (a belief that necessarily disallows the possibility of separation). So I will attempt to outline the issue as I understand it so that we may have a common ground for discussion.
When people talk about the need to remove the strategic vote, what they mean is they want to remove or reduce the circumstances under which a voter must choose between voting for the candidate they personally favor (their sincere choice) and a candidate they think might win (their strategic choice). Such situations are common when using First-Past-The-Post (A.K.A. plurality voting) methods where voters casting ballots for their sincere choice often feel that their vote is "being thrown away" if it's going towards a candidate who has no reasonable chance of winning. And so instead they often cast their ballot strategically for a less desirable candidate who is more likely to win. This in turn leads to a less representative government, which is bad, because representative government is the goal of democracy! It's because of this that advocacy of other voting methods that offer no incentive to vote strategically is gradually increasing.
Want to learn more about different voting methods? This isn't a bad place to start: http://en.wikipedia.org/wiki/Voting_system
That is NOT what Voter Verifiable Voting means. VVV simply means that the voter can personally verify that their vote was recorded properly.
Today, with paper ballots, this is as easy as looking at your ballot and making sure you put the mark next to the intended candidate. You don't keep this with you when you leave the polling station; you drop it in the ballot box. Nobody knows how you voted but you.
Today, using electronic voting machines, this does not exist. The machine may tell you that it cast your vote for the candidate you intended, but ultimately the machine records your vote on a memory card that you don't have access to. In other words, the voter has substituted their own verification of the vote for the machine's verification of their vote. And without the voter being able to personally verify that their intended candidate received their vote, in the end, they're forced to trust that the machine has made no error and his recorded the vote honestly. Many voters do not have this trust. And with the myriad accounts of voting machine errors and tampering in circulation today, rightly so.
So when people loudly decry that our voting machines need Voter Verifiable Voting, what they're usually looking for is a paper trail that they can personally see and understand. For example, a slip of paper that says "You voted for John Smith for President". This paper receipt need not be kept by the voter! It could be verified by the voter and then dropped in a secure ballot receipt box. The voter need not even personally touch this receipt! It could be presented behind glass for verification by the voter, to be secured upon the voter's approval. And while the machine would still be used to count and report the votes, the paper receipts would allow for meaningful recounts and would provide a paper trail for audits. These are good, important things.
And while there exists a poorly informed minorty that would like to be able keep this voter verifiable receipt with them when they leave the polls, these people represent a distinct subset of people calling for Voter Verifiable Voting in general. So let's be sure not write off VVV as a whole because there exists a handful of people who want to implement it poorly.
While the Neuros seems to have a lot of cool features, it would appear that Mac users need not apply. Here are the system requirements as taken from their website, for those who care:
OS: Microsoft® Windows 98SE/Me/2000/XP
CPU: Pentium 233MHz or higher
Memory: 64MB minimum
Hard Drive: 160MB
USB Port